AI: The Pentagon bans Anthropic from contracts and Amodei denounces a punitive decision

The CEO of Anthropic, Dario Amodei, has just stepped up against an unprecedented decision by the Trump administration. Washington has designated his company as a “risk to the supply chain” of American defense, opening the way to an unprecedented legal battle between a major American tech firm and its own government.

A sudden break after eighteen months of collaboration

It all starts in 2024 with a $200 million contract signed between Anthropic and the Department of Defense. At the time, the company was proud to be the first advanced AI company to operate on classified US military networks. A source of pride. A strong commercial argument.

Eighteen months later, this same contract is at the heart of an open legal confrontation with Washington.

The sticking point is simple: the Pentagon wanted unrestricted access to Claude, Anthropic’s flagship model. 

The company, for its part, set two non-negotiable conditions, no integration into fully autonomous weapons systems, no mass domestic surveillance of American citizens. Months of private negotiations changed nothing. The ultimatum set at 5:01 PM Friday passed without agreement. Anthropic did not move.

The administration’s response was immediate and brutal. Pete Hegseth designated Anthropic as a “national security risk in supply chain matters,” a label usually reserved for foreign companies under hostile influence, such as Huawei or ZTE. 

Several lawyers have already called this decision a dangerous precedent. Trump then hammered the point on Truth Social, in uppercase: all federal agencies must “IMMEDIATELY CEASE all use of Anthropic technology.”

Anthropic stands firm, OpenAI benefits

In front of CBS News cameras, Dario Amodei conceded nothing. Calling the Pentagon’s decision “unprecedented” and “punitive,” Anthropic’s CEO firmly reaffirmed the core of his position:

These are fundamental things for Americans: the right not to be spied on by the government, the right for our military officers to make decisions about war themselves, and not to entrust them entirely to a machine.

He specified that he was not opposed on principle to automated weapons, but that current AI models are simply not reliable enough to operate without human supervision in a lethal context. He also called on Congress to quickly legislate to regulate AI usage in national surveillance programs.

Anthropic announced that it will contest the designation in court, relying on article 3252 of title 10 of the US Code, which legally limits the scope of this designation to Department of Defense contracts only.

Meanwhile, OpenAI seized the opportunity. Just hours after Hegseth’s announcement, Sam Altman confirmed signing an agreement to deploy OpenAI models on US military networks. The online reaction was strong, with many seeing it as an endorsement of mass surveillance and AI militarization.

Anthropic chose its principles over its contracts. A risky bet, but consistent with its mission stated since its origins: to develop a safe AI. The upcoming legal battle will show whether this stance is tenable against the American administrative machine.